EAGER: Protecting Election Integrity Via Automated Ballot Usability Evaluation

任何导致计票结果与选民意图不同的事情都是对选举完整性的威胁。虽然对选举完整性的大多数威胁都与安全有关，但对选举完整性的另一个关键威胁是可用性。当选民由于糟糕的选票设计而无法成功地传达他们的意图时，无论安全水平如何，这都会威胁到选举的完整性。传统的可用性测试方法不能很好地扩展到美国各地每次选举中部署的成千上万种不同的选票样式，因此需要一种替代解决方案。本研究旨在通过发展必要的科学来解决这个问题，以支持一种工具，当将选票作为输入时，该工具会对该选票是否可能导致选民错误进行评估，如果有的话，选票上最有可能出现这些错误的地方。这项研究是基于一个成熟的认知架构开发的计算人类表现模型。这种架构在过去已经被许多研究人员成功地应用于其他可用性问题，尽管从未应用于投票。在理解可视化分组的领域，将需要对现有建模系统进行扩展。此外，该系统将以一种新颖的方式使用。大多数类似的问题已经通过构建一个代表单一策略的单一人类模型来解决。在这个项目中，研究人员正在构建一系列基于对选票完成空间和选民可用的视觉搜索策略的探索的模型。然后，在策略空间的每个点重复运行随机模型，以发现哪些选民策略和选票设计的交集导致高错误率。研究人员正在使用现有已知的不良选票以及新的行为数据来验证这种方法。这项研究的结果将允许构建一个选票分析工具，选举官员可以使用它在选举日部署之前识别可能存在问题的选票。

Anything that causes the vote tally to differ from the intent of the voters is a threat to election integrity. While most threats to election integrity have concerned security, there is another critical threat to election integrity: usability. When voters are unable to successfully communicate their intent due to poor ballot design, this threatens the integrity of the election, no matter what the level of security is. Traditional usability testing methods do not scale well to the tens of thousands of different ballot styles deployed across the United States in each election, so an alternative solution is necessary. This research aims to address this problem by developing the science necessary to support a tool that, when given a ballot as input, produces an assessment of whether or not that ballot is likely to lead to voter error, and if so, where on the ballot these errors are most likely to occur.This research is based on computational human performance models developed with a well-established cognitive architecture. This architecture has been successfully applied to other usability problems by numerous researchers in the past, though never to voting. Extensions to the existing modeling system will be required in the domain of understanding visual grouping. In addition, the system will be used in a novel way. Most similar problems have been addressed by constructing a single human model that represents a single strategy. In this project, the researchers are constructing a family of models based on an exploration of the space of ballot completion and visual search strategies available to voters. Then, the stochastic model is run repeatedly at every point in the strategy space in order to discover which intersections of voter strategies and ballot designs lead to high error rates. The researchers are validating the approach using existing known bad ballots as well as with new behavioral data. The results of this research will allow the construction of a ballot analysis tool that could be used by election officials to identify potentially problematic ballots before deploying them on election day.